Neuromechanism Study of Insect–Machine Interface: Flight Control by Neural Electrical Stimulation

Huixia Zhao1, Nenggan Zheng2*, Willi A. Ribi3, Huoqing Zheng1, Lei Xue2,4, Fan Gong2,4, Xiaoxiang Zheng2,4,5, Fuliang Hu1* 1 College of Animal Sciences, Zhejiang University, Hangzhou, China, 2 Qiushi Academy for Advanced Studies (QAAS), Zhejiang University, Hangzhou, China, 3 The Private University of Liechtenstein, Dorfstrasse 24, Triesen, Liechtenstein, 4 Department of Biomedical Engineering, Zhejiang University, Hangzhou, China, 5 Key Laboratory of Biomedical Engineering of Ministry of Education, Zhejiang University, Hangzhou, China

Abstract The insect–machine interface (IMI) is a novel approach developed for man-made air vehicles, which directly controls insect flight by either neuromuscular or neural stimulation. In our previous study of IMI, we induced flight initiation and cessation reproducibly in restrained honeybees (Apis mellifera L.) via electrical stimulation of the bilateral optic lobes. To explore the neuromechanism underlying IMI, we applied electrical stimulation to seven subregions of the honeybee brain with the aid of a new method for localizing brain regions. Results showed that the success rate for initiating honeybee flight decreased in the order: a-lobe (or b-lobe), ellipsoid body, lobula, medulla and antennal lobe. Based on a comparison with other neurobiological studies in honeybees, we propose that there is a cluster of descending neurons in the honeybee brain that transmits neural excitation from stimulated brain areas to the thoracic ganglia, leading to flight behavior. This neural circuit may involve the higher-order integration center, the primary visual processing center and the suboesophageal ganglion, which is also associated with a possible learning and memory pathway. By pharmacologically manipulating the electrically stimulated honeybee brain, we have shown that octopamine, rather than dopamine, serotonin and acetylcholine, plays a part in the circuit underlying electrically elicited honeybee flight. Our study presents a new brain stimulation protocol for the honeybee–machine interface and has solved one of the questions with regard to understanding which functional divisions of the insect brain participate in flight control. It will support further studies to uncover the involved neurons inside specific brain areas and to test the hypothesized involvement of a visual learning and memory pathway in IMI flight control.

Citation: Zhao H, Zheng N, Ribi WA, Zheng H, Xue L, et al. (2014) Neuromechanism Study of Insect–Machine Interface: Flight Control by Neural Electrical Stimulation. PLoS ONE 9(11): e113012. doi:10.1371/journal.pone.0113012 Editor: Paul Graham, University of Sussex, United Kingdom Received May 11, 2014; Accepted October 22, 2014; Published November 19, 2014 Copyright: ß 2014 Zhao et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: The authors confirm that all data underlying the findings are fully available without restriction. All relevant data are within the paper and its Supporting Information files. Funding: The author FLH received the funding from the National Natural Science Foundation of China (No. 61031002), and the Modern Agroindustry Technology Research System from the Ministry of Agriculture of China (CARS-45). The author XXZ received the funding from the National Natural Science Foundation of China (No. 61003150), and the Fundamental Research Funds for the Central Universities. This work is partially supported by Zhejiang Provincial Natural Science Foundation of China (LZ14F020002). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * Email: [email protected] (NZ); [email protected] (FH)

Introduction behavior in locusts by a chronic electrical stimulation method on different brain sites [10]. And Blondeau (1981) had stimulated Insects, due to their impressive flight skills, are ranked among neurons in the lobula plate of free-moving and fixed Calliphora the best models for studying mechanisms of flight control, and for erythrocephala to evoke course control [11]. In these studies, use in the development of biomimetic micro air vehicles (MAVs) electrical stimulation was used as one of the tools of neuroethology [1]. As MAVs have become an increasingly hot topic of research, to investigate the relationship between animal behavior and the great advances have been made in the past few decades, but nervous system. Beyond that, electrical stimulation was also used significant challenges still remain with regards to payload mass, to artificially control the locomotion of an autonomous bio-robotic flight range, and speed [2,3]. A novel approach called the insect– system. Holzer and Shimoyama (1997) had induced the escape machine interface (IMI), which directly controls the flight behavior turn of cockroach via electrical stimulation to antennae, and built of insects by either neuromuscular or neural stimulation, has been an electronic backpack to control cockroach walking [12]. developed in recent years and promises to overcome some of the As to the IMI for MAVs and flight control, researchers challenges facing MAVs [4,5]. managed to elicit flight initiation and cessation in beetles (Cotinis The use of electrical stimulation to induce behavior in insects is texana and Mecynorhina ugandensis) by applying electrical not new. Singing behavior of the cricket and grasshopper had been stimulation through two electrodes implanted in the bilateral elicited by electrical stimulation of the brain [6,7] and descending optic lobes [13,14,15,16]. Their experiment was inspired by the fibres [8,9]. Rowell (1963) had produced various activities immediate flight cessation of untethered Mecynorhina ugandensis including antennal movements, locomotion, feeding, and sexual in response to abrupt darkening of the environment, which led the

PLOS ONE | www.plosone.org 1 November 2014 | Volume 9 | Issue 11 | e113012 Flight Control in Honeybees researchers to hypothesize that the optic lobes might strongly Millonig’s buffer (pH 7.2) for 2–4 h before postfixation in a modulate flight initiation and cessation [5]. While alternating buffered 2% OsO4 solution for 1 h at room temperature. They positive and negative potential pulses at 100 Hz (20% duty cycle) were then dehydrated in an ethanol series (50%, 70%, 80%, 95%, initiated wing oscillations, a relatively long duration pulse caused 10 minutes each and 26100%, 15 minutes each), cleared twice in wing oscillations to cease [16]. Another study found that propylene oxide (15 and 20 minutes), and placed back into stimulating the antennal lobes with 20 Hz, 3.5 Vpp pulses initiated propylene oxide and Araldite mixture (1:1, 2–3 h) before being flight, while 50 Hz, 3.5 Vpp pulses stopped flight, in Manduca embedded in Araldite. Consecutive 1 mm thick sections were cut sexta [17]. In addition, ‘‘throttling’’ (frequency and stroke on a rotatory microtome using a diamond knife and stained with amplitude of wing oscillation) and turning modulation of insect toluidine blue. flight had been achieved by stimulating the optic lobes and basalar muscles [13,16], neck muscles [17] or ventral nerve cord Fixation and positioning system [18,19,20]. One of our previous studies developed a stimulation To assist with positioning the microelectrodes into specific brain protocol in restrained honeybees (Apis mellifera), using alternating subregions, we set up a fixation and positioning system by positive and negative electrical pulses (4 Vpp-40 Hz-40% duty assembling several pieces of equipment, including a honeybee cycle, t = 5 ms, m = 30) between two microwire electrodes clamping device (manufactured by our team), a digital stereotaxic implanted into the bilateral optic lobes to reproducibly generate apparatus (RWD Life Science Co., China), and a gimbaling flight initiation and cessation [21]. stereomicroscope with cold light illuminator (RWD Life Science It is difficult to elucidate the neuronal mechanisms of IMI from Co., China). The set-up is shown in Fig. 1. previous studies because identification of these mechanisms Honeybees were cold anesthetized before being fixed onto the requires electrophysiological or optical recordings of neuronal clamping device. A stereotyped procedure was applied to ensure activity in insects during stimulation [16]. We made some consistent head position of each fixed honeybee. When the rostral preliminary attempts to record neuronal activity, but unfortunately head was pressed down, the interspace between the retral head encountered problems due to interference from noise generated by and the clamp plates was filled up with beeswax–rosin mixture electrical stimulation of the brain (i.e. stimulus artifact) and robust (2:1). Meanwhile, special attention was paid to adjust the head avoidance behaviors of the insects (such as abdomen twitch and midline in accordance with a reference line on the clamp plates wing oscillations). We therefore posed an alternative question, (Fig. 1). After fixation, the honeybee underwent limb amputation namely how will stimulation of specific brain subregions (with the because legs gripping onto the clamp plates would hinder wing exception of the optic lobe) affect flight initiation? Solving this movement. Antennae were removed to facilitate electrode question will narrow the search for the neurons involved in IMI to implantation. Then, under the stereomicroscope, a rectangular specific brain subregions; by comparing our data with other window was cut in the cuticle between two compound eyes and honeybee neurobiological studies it may be possible to deduce the between antennae and ocelli using a scalpel with a carbon steel neural pathways involved. blade. After removing the glands, membrane and tracheae that To accomplish this aim, we first established a new method for cover the brain from the front, intact brain was exposed and localizing brain regions to help with embedding microelectrodes perfused with bee saline to keep it moist (Ringer’s solution: 130 reproducibly into targeted brain subregions. Then, we applied an NaCl, 6 KCl, 4 MgCl2, 5 CaCl2, 160 sucrose, 25 glucose, 10 innovative stimulation protocol to the unilateral brain to study the HEPES, in mM; pH 6.7, 500 m Osmol). effects of stimulating different subregions on flight initiation. Lastly, we adopted the method normally used by others to study Electrode implantation the effects of neurotransmitters, neuromodulators and neurohor- A new method for locating brain regions was developed to allow mones such as biogenic amines on flight control. implantation of the stimulating electrode into a specific brain subregion. Positional data obtained from frontal and horizontal Materials and Methods brain sections were used as references for medio-lateral and antero-posterior (depth) localization of brain subregions respec- Animals tively. Brain surface landmarks discovered through stereomicro- Worker honeybees (Apis mellifera L.) were captured at the scopy and light microscopy were used as references for dorso- entrance to hives in Zhejiang University and maintained in a ventral localization. perforated bottle containing sugar and wet paper. Throughout the The stimulating electrodes are Elgiloy–stainless steel microelec- experiments, honeybees were kept warm under a heat lamp (in trodes, which have Parylene insulation until approximately 66 mm winter) and regularly supplied with new wet paper. from the tip end (impedance 0.5 MV) and 1–2 mm tip diameter (World Precision Instruments, USA). After a stimulating electrode Brain microtomy had been embedded in a specific brain subregion, an indifferent Paraffin-embedded thick sections: brains were dissected out of electrode made from formvar-insulated nichrome wire (bare head capsules, cleaned to remove fat and tracheae, and fixed in diameter 50.8 mm, A–M Systems, USA) was placed in bee saline 4% formaldehyde overnight. They were then dehydrated in an outside the brain surface to complete the current return path. increasing ethanol series (50%, 70%, 80%, 2695%, 26100%, 15 minutes each), cleared in dimethylbenzene (2610 minutes), Electrical stimulation infiltrated with soft wax (260.5 h) and hard wax (1 h) before A 0.3 s stimulus train of rectangular biphasic (1 ms each phase) being embedded in hard wax. Consecutive 4 mm thick sections pulses at 200 Hz was generated by an Isolated Pulse Stimulator were cut on a rotary microtome (HM 325, MICROM, Germany) (model 2100, A–M Systems, USA) and used for stimulation. The using solid razor blades, and were then stained with hematoxylin- isolated constant current was set to two amplitudes: 10 mA as low eosin before being mounted in neutral balsam. intensity and 30 mA as high intensity (Fig. 2). Araldite-embedded semithin sections: dissected honeybee brains Stimulation experiments were carried out in 40 and 100 were fixed in a primary fixative containing 2% paraformaldehyde, honeybees using low and high current intensity, respectively. Both 2.5% glutaraldehyde and 2.2 g sucrose per 100 ml solution in experiments were divided equally into two groups in which seven

PLOS ONE | www.plosone.org 2 November 2014 | Volume 9 | Issue 11 | e113012 Flight Control in Honeybees

Figure 1. Equipment set-up, honeybee clamping device, and head fixation (the red dotted line indicates head midline). doi:10.1371/journal.pone.0113012.g001 brain subregions were stimulated in opposite sequences to 1022 M before they were diluted into 1025 M solutions. For each eliminate excitement interactions. Stimulation of each brain drug group, 5 ml1025 M solution was dripped onto the brain subregion was carried out according to the rule that the next surface of ten honeybees using a 50 ml syringe. Electrical stimulus could be given after the most recent stimulus failed to stimulation was performed 5 minutes after drug application. induce flight, and stimulation would be stopped when flight was induced or three stimuli were given. That is consecutive stimuli are Behavior recording used following this rule. A digital camera (HDR-SR12E, Carl Zeiss lens Vario-Sonnar In the pharmacological manipulation experiments, stimulation T*, SONY) was positioned in front of the restrained honeybee and for all six drug groups was carried out on the a-lobe under low used to record behaviors in response to electrical stimulation. In intensity conditions. Videos were made at a specific time point neurotransmitter experiments, three videos were taken at 5-minute before or after drug delivery (see below Behavior recording), intervals before drug delivery, and four videos were taken 5, 10, 15 during which honeybee responses to three segregated stimuli were and 20 minutes after drug delivery. examined. Data analysis Prussian blue verification The initiation and termination of flight were distinguished by After stimulation, the Isolated Pulse Stimulator was modulated high wing-beat frequency from video frames displayed for 1/25 to generate a dissociating pulse (20 mA DC, 15–20 s), which second using the software Corel VideoStudio Pro X5 (Ulead 3+ partially dissociated Fe from the stimulating electrode and Information Inc., USA) (see Figure S1). deposited it in surrounding brain tissue. Brains were then In the neurotransmitter experiments, flight duration due to each sectioned in paraffin and dyed with a mixture of potassium stimulus was calculated from videos. To verify that the honeybee ferrocyanide and hydrochloric acid solution to generate prussian was showing stable performance before application of the drug, we blue at the stimulation site. performed pairwise comparisons of flight durations in the three pre-delivery videos. When there were no significant differences, Drug dispensing the three pre-delivery videos were considered as an ensemble to Three biogenic amines (dopamine hydrochloride, (6)-octopa- compare with each video taken after drug delivery in order to mine hydrochloride, serotonin hydrochloride), acetylcholine chlo- investigate the time-varying effect of drug delivery on flight ride, phentolamine hydrochloride (octopamine receptor antago- duration. Lastly, four post-delivery videos were taken as an nist) and (+)-butaclamol hydrochloride (dopamine receptor ensemble to compare with the pre-delivery ensemble. All antagonist) (all chemicals were purchased from Sigma-Aldrich comparisons were made using non-parametric Mann–Whitney Ltd.) were dissolved separately in bee saline at a concentration of U tests in SPSS statistics 19.0 software. Significant differences were

Figure 2. Electrical pulses for stimulation. doi:10.1371/journal.pone.0113012.g002

PLOS ONE | www.plosone.org 3 November 2014 | Volume 9 | Issue 11 | e113012 Flight Control in Honeybees

Figure 3. Brain sections and measurement of medio-lateral and antero-posterior positional data. (A) Frontal paraffin thick section showing the a-lobe, mushroom body calyx, and antennal lobe. (B) Frontal Araldite semithin section showing the b-lobe, ellipsoid body, medulla, lobula and dorsal lobe. (C) Horizontal Araldite semithin section showing the a-lobe, b-lobe, ellipsoid body, medulla and lobula. Scale bar = 100 mmin all pictures. doi:10.1371/journal.pone.0113012.g003 accepted at p,0.05. A group-level analysis was also performed brain (see Figure S2). Because little is known about the distribution using ANOVA in SPSS. of data for brain subregion positions, we chose the non-parametric Kruskal–Wallis H test to compare the data from different Results honeybees. Results showed no significant differences in both medio-lateral and antero-posterior positional data for five brain Positional data subregions (p.0.05, Table 1). And using non-parametric Mann– Positional data for five brain subregions (a-lobe, ellipsoid body, Whitney U test to compare the medio-lateral positional data of five lobula, medulla, antennal lobe) in two directions, medio-lateral brain subregions obtained by two microtomy techniques, we had and antero-posterior, were obtained from frontal sections of six not found significant differences (p.0.05, Table 2). Mean values brains and horizontal sections of three brains. Two microtomy for both medio-lateral and antero-posterior positional data were techniques (see Methods) were used to eliminate methodical error. calculated and used to assist electrode implantation (Table 3 and Eleven slices from serial brain sections of each honeybee were Table 4). selected and photographed before being measured. The medio- lateral positional data for five brain subregions were measured as Discovery of new brain landmarks perpendicular distance to the brain midline (Fig. 3A, B, C), and We identified some new brain landmarks by stereomicroscopy antero-posterior positional data were measured as perpendicular and light microscopy. Under the stereomicroscope with bilateral depth to the brain surface (Fig. 3C). All data were recorded in cold light illumination, two small mango-shaped regions could be Microsoft Excel and processed into curve charts to show the distinguished from their surroundings on the central brain as they positional distribution of different subregions in the honeybee were distributed symmetrically along the brain midline and

Table 1. Comparisons (Kruskal–Wallis H test) of the medio-lateral (N = 6) and antero-posterior (N = 3) positional data obtained from different honeybees.

brain subregions a-lobe ellipsoid body lobula medulla antenal lobe

p value (medio-lateral) 0.327 0.111 0.402 0.637 0.227 p value (antero-posterior) 0.618 0.149 0.088 0.096 0.165

doi:10.1371/journal.pone.0113012.t001

PLOS ONE | www.plosone.org 4 November 2014 | Volume 9 | Issue 11 | e113012 Flight Control in Honeybees

Table 2. Comparisons (Mann–Whitney U test) between the medio-lateral positional data of five brain subregions obtained by two microtomy techniques.

brain subregions a-lobe ellipsoid body lobula medulla antennal lobe

p value 0.347 0.561 0.119 0.261 0.098

doi:10.1371/journal.pone.0113012.t002 emitted dark aqueous reflection (not obvious under a centralized effective. Of 100 successful cases, 76 honeybees’ flight was initiated light source, Fig. 4A). Prussian blue verification of an electrode by one stimulus, and the others by two or three stimuli (Fig. 5B). implanted through the mango-shaped region demonstrated that Stimulation of the medulla and lobula failed to induce flight this area was the a-lobe (Fig. 4C). The two antennal lobes were initiation in 40 honeybees tested under low current intensity easily identified under the stereomicroscope from their position conditions, only inducing lifting of the wings without flapping ventral to the protocerebrum and dorsal to the antennal roots (Fig. 5A). But when high intensity current was used to stimulate (Fig. 4A). the medulla or the lobula, flight was induced in 43% and 81% of Light microscopy of serial horizontal brain sections revealed honeybees, respectively (N = 100, Fig. 5B). Of 43 successful cases that the ellipsoid body lies on the brain midline and between the of medulla stimulation, 18 honeybees initiated flight in response to upper borders of the two a-lobes (Fig. 4E). The mango-shaped a- three consecutive stimuli, 20 with two stimuli and 5 with only one lobe landmarks can thus be used to localize the ellipsoid body too. stimulus. For the lobula, successful flight of 17 honeybees were Microscopy of serial frontal brain sections indicated that the arc- induced by three consecutive stimuli, 24 by two stimuli and 40 by shaped border of the gap between the two antennal lobes was only one stimulus. aligned approximately with the centers of the medulla and lobula Stimulation of the antennal lobe had no effect on flight (Fig. 4B, F, G). The arc-shaped border can thus be used as a initiation, but induced folding wings or slight lifting of the wings landmark for localizing the medulla and lobula. Microscopy of (Fig. 5A, B). However, stimulation of the deeper subesophageal serial frontal brain sections also indicated that the subesophageal ganglion (electrode depth .300 mm) with high current intensity ganglion was posterior to the antennal lobes. Prussian blue successfully induced flight initiation. verification of an electrode implanted deep through the antennal To gain a detailed impression of honeybee behaviors in lobe demonstrated that it reached the subesophageal ganglion response to stimulation of specific brain subregions, readers are (Fig. 4H, I). Measurements in serial horizontal brain sections referred to Video S1 and Video S2. demonstrated that the a-lobe medio-lateral positional data (214693 mm) was in the range of b-lobe medio-lateral positional Effects of biogenic amines or receptor antagonists on data (1806180 mm). Therefore the location of b-lobe could be flight activity determined by the a-lobe landmark on surface. Moreover, the a- For six drug groups (60 honeybees in total), flight duration in lobe had a shallow position (95695 mm) compared to the deep b- response to stimulation was tested before drug delivery, and each lobe (284685 mm) (Fig. 3C). honeybee was tested nine times, i.e., 540 tests in total. Flight duration varied enormously between individuals (0.23–19.45 s), Effects of stimulating different brain subregions on flight with an average of 1.8 s. However, comparisons of the nine flight initiation durations recorded from each honeybee did not reveal significant differences (p.0.05, N = 60), which means that all honeybees were By implanting the electrode through the a-lobe landmark at showing stable performance before they received a drug. different depths, we achieved stimulation of either the a-lobe or Whisker diagrams were used to display the analysis results of b the -lobe (Fig. 4C, D). While an electrode implanted at a depth flight durations tested before and after drug delivery (see Fig. 6). a b less than 180 mm stimulates the -lobe, the -lobe is stimulated by Each diagram shows only one of the ten honeybees tested in a an electrode implanted at a depth of 180–360 mm. Stimulation of drug group. The mean value of flight duration in the pre-delivery the a-lobe consistently triggered flight initiation with a single video ensemble or each post-delivery video is shown in the middle stimulus in both low and high intensity groups (success rate 100%, as a short horizontal bar with a cross. The higher and lower N = 40, 100) (Fig. 5A, B). Stimulation of the b-lobe had the same squares on each vertical line represent the longest and shortest effect on inducing honeybee flight (Fig. 5A, B). We also briefly flight durations in that video or video ensemble respectively. tested stimulation of the mushroom body calyx, but this failed to Moreover, a box and whisker plot was used to summarize the induce honeybee flight even with a higher current intensity. group averaged data and to show the group differences in flight In the low intensity group, however, stimulation of the ellipsoid duration for the different drug types (Fig. 7). body required two or three consecutive stimuli to initiate flight in No significant differences were seen in the dopamine (DA), DA- all honeybees (100%, N = 40) (Fig. 5A). But under high current receptor antagonist (butaclamol, Bu), serotonin (5-HT) and intensity conditions, stimulation of the ellipsoid body was more acetylcholine (Ach) experiments (p.0.05, N = 40, Fig. 6iii, iv,v,vi).

Table 3. Medio-lateral positional data [mm] for five brain subregions.

compartment a-lobe ellipsoid body lobula medulla antennal lobe

positioning data 214 (±93) 0 (±70) 649 (±116) 906(±138) 250 (±155)

doi:10.1371/journal.pone.0113012.t003

PLOS ONE | www.plosone.org 5 November 2014 | Volume 9 | Issue 11 | e113012 Flight Control in Honeybees

Table 4. Antero-posterior positional data [mm] for five brain subregions.

compartment a-lobe ellipsoid body lobula medulla antennal lobe

positioning data 95 (±95) 326 (±47) 293 (±111) 205 (±205) 155 (±155)

doi:10.1371/journal.pone.0113012.t004

However, significant effects were seen with octopamine (OA) and Discussion its receptor antagonist (phentolamine, Ph). Ten honeybees in the OA group all significantly increased their flight duration in The method of combining brain landmarks with positional data response to having 5610211 mol OA dripped onto their brain has been proven adequate for the localization of specific brain surface (p,0.05, N = 10). Comparisons of four single videos subregions in honeybees. It was possible to reproducibly localize obtained after delivery with the pre-delivery video ensemble seven brain subregions. Although this method is not accurate revealed a delayed flight elongation effect in eight honeybees enough to locate the position of specific neurons, it has already (Fig. 6i). Significant differences were also obtained when compar- allowed great progress to be made in localizing brain regions in ing the pre-delivery video ensemble with the post-delivery video vivo. Previous studies have usually used landmarks on the brain ensemble (p,0.05, N = 10, Fig. 6i). By contrast, ten honeybees surface as references to localize brain subregions, and in tested after having 5610211 mol Ph dripped onto their brain combination with recording or stimulation results to localize surface all showed significantly shorter flight duration, and indeed neurons [22]. For example, the lobula plate in the fly brain can be flight was sometimes even suppressed completely. Comparisons identified by a characteristic branching pattern of silvery trachea between the pre-delivery and post-delivery video ensembles that covers its posterior surface [23]. However, established revealed significant differences in all honeybees (p,0.05, N = 10, landmarks for most brain subregions in many model animals Fig. 6ii), but no significant differences were obtained by compar- (e.g. the honeybee) are still limited. Plus, for many brain ing the pre-delivery video ensemble with four single post-delivery subregions, we do not know what the effects of stimulation will videos (p.0.05, N = 10). be; in fact, these are the problems we wish to solve. From group-level analysis, we found statistical differences in The success rate for flight initiation is higher with electrical both OA and Ph groups which are indicated by stars in Fig. 7. But stimulation of the unilateral a-lobe (or b-lobe) than with there was also one star above the box and whisker of 5-HT in the stimulation of the ellipsoid body, which is turn higher than with 10 min group. It might be caused by the high variability of flight stimulation of the lobula and the medulla, respectively. Other performance among individuals. regions such as the antennal lobe and the protocerebrum area

Figure 4. New brain landmarks and prussian blue paraffin sections verifying the stimulation sites. Scale bars 50 mm. (A) Honeybee brain under the stereomicroscope. The two pink arrows indicate landmarks for the a-lobes: small mango-shaped regions. The two stars indicate the antennal lobes. (B) Frontal brain section showing the arc-shaped border of the gap between the two antennal lobes, which is aligned approximately to the centers of the medulla and lobula. (C–I) Paraffin-embedded frontal brain sections dyed with prussian blue. Arrows show the stimulation sites on the a-lobe, b-lobe, ellipsoid body, lobula, medulla, antennal body and suboesophageal ganglion, respectively. Note that each picture shows a hole of around 15 mm with a large surrounding sphere with different staining; this arose because the monophasic dissociating pulse (20 mA DC, 15–20 s) generated polarization and hydrolysis around the electrode tip and produced tissue damage. doi:10.1371/journal.pone.0113012.g004

PLOS ONE | www.plosone.org 6 November 2014 | Volume 9 | Issue 11 | e113012 Flight Control in Honeybees

Figure 5. Success rates of flight initiation in response to stimulation of six brain subregions. (A) Low intensity group. (B) High intensity group. doi:10.1371/journal.pone.0113012.g005 mushroom body calyx (prussian blue verified) failed to induce the stimulation directly depolarizes the ‘‘giant fiber’’ motor honeybee flight when stimulated using the same protocol. neurons which connect the insect brain to the flight muscles and Tehovnik (1996) stated that the total number of neurons activated mediate the escape flight. Alternatively, the stimulation might directly by a given current was not only dependent on the current depolarize sensory afferents to the brain that lead to alteration of intensity but also on the excitability of the neurons [24]. Indeed, the flight central pattern generator. However, the homologous when the same brain subregion was stimulated in honeybees in our giant fibers haven’t been found in honeybees. And with a number experiments, the high intensity group (30 mA) always resulted in a of other armaments to defend themselves, why would honeybees higher success rate than the low intensity group (10 mA). Given need to escape with such a specialized and hard-wired system? that electrode placement is not sufficiently accurate to target a Whereas, studies had found a number of descending neurones specific neuron, we assume that flight initiation is reproducibly (DNs) in honeybees which receive visual inputs from the optic induced in different honeybees by the excitation of different lobes and descend to the thoracic ganglia to control flight course groups of neurons within a brain subregion. Accordingly, we have [25,26]. All of these neurons have their major dendritic demonstrated that multiple subregions in the insect brain, rather arborizations located in the posterior deutocerebrum and imme- than just the optic lobes, can be used to manipulate flight diately lateral to the oesophageal foramen [25,27]. Some neurons initiation, but we are not sure if these different subregions have give off branches in the tritocerebrum and suboesophageal different functions with regards to the exquisite ‘‘throttling’’ and ganglion [26]. Our stimulation of the suboesophageal ganglion turning modulation of flight. had induced flight initiation in honeybees. Also, flight was initiated In their study of beetle flight control, Sato et al. hypothesized by stimulation of two optic neuropils — the lobula and the two possible neural pathways for the flight initiated by electrical medulla, respectively, in some cases. These behavioral responses pulses applied between bilateral optic lobes [16]. One possibility is might realize through the depolarization of such DNs. To make

PLOS ONE | www.plosone.org 7 November 2014 | Volume 9 | Issue 11 | e113012 Flight Control in Honeybees

Figure 6. Whisker diagrams showing analysis results. The drug used is abbreviated at the top of each diagram. Stars in i indicate that the flight durations from three post-delivery videos are all statistically different from the pre-delivery video ensemble (p,0.05). Two stars above the transverse line in ii and ii indicate statistical differences in flight durations between two video ensembles (before and after delivery) (p,0.05). doi:10.1371/journal.pone.0113012.g006 sure the involvement of such DNs in the electrically elicited flight the LTP in honeybee brain by electrically stimulating Kenyon cells initiation and to make clear how these DNs control the flight at high frequency (100 Hz) and recording from a single, identified muscles, it awaits for the further study of neural recordings. mushroom body output neuron, the PE1 [29]. Both studies An interesting potentiation effect of consecutive stimuli was contribute to the mechanism of olfactory learning and memory observed in some stimulation experiments of the ellipsoid body consolidation. We stimulated the ellipsoid body or optic lobe by and optic lobes. Where a single stimulus could only elicit shock consecutive stimuli which have a high frequency (200 Hz) and use responses (shaking legs, abdomen twitch, sting reflex, lifting wings a time interval of 2–3 seconds between bouts of stimuli. It seems without flapping), two or three consecutive stimuli induced flight somewhat likely that our stimulation might induces the LTP in the initiation. This effect might have arisen due to the long-term neural pathway mediating flight initiation. potentiation (LTP). Oleskevich et al. (1997) had reported the first In addition, we had found another surprising phenomenon demonstration of long-term synaptic plasticity by long-lasting when we stimulated the a-lobe first with five or more consecutive potentiation in honeybee brain [28]. They recorded the extracel- stimuli and then stimulated other brain subregions. We found the lular field response of mushroom body Kenyon cells after electrical stimulation of any other brain subregion could induce flight stimulation of the olfactory input pathway. LTP was induced by initiation by only one stimulus (personal observation, not included low-frequency stimulation (0.02–1 Hz). Another study discovered in Results). This phenomenon were held for longer than 10

PLOS ONE | www.plosone.org 8 November 2014 | Volume 9 | Issue 11 | e113012 Flight Control in Honeybees

Figure 7. A box and whisker plot showing group-level data analysis, in which the box shows the mean and the whisker shows the standard error (SE). Stars indicate statistical differences that are found in the comparison of group averaged flight durations in pre-delivery ensemble and each post-delivery video by ANOVA. doi:10.1371/journal.pone.0113012.g007 minutes generally. It is believed that the mushroom bodies of the honeybee flight. Octopamine applied to the honeybee brain insect brain are higher-order sensory integration centers that are increased flight duration; however, application of an octopamine– involved in the memory formation and storage [29]. Together receptor antagonist blocked flight induction or reduced flight with the potentiation effect we have presented above, it is very duration. These results are different from those of others by likely that our brain stimulation have evoked flight behavior in negating the effect of dopamine and serotonin on flight honeybees through a descending pathway which also associates modulation (see [36]). Perhaps it is because we have induced with a visual learning and memory pathway. flight behavior by direct brain stimulation and excited a specific It is generally agreed that octopamine can act as a key neural circuit, of which a key element is octopamine. But we won’t neuromodulator as well as a neurohormone to modulate the insect deny the possibility that octopamine might play a role by acting on flight [30]. Octopamine can alter the plateau potential and the central pattern generator (i.e. thoracic ganglia) and flight excitability of the interneurons in flight central pattern generator muscles. Further studies using the picospritzer to limit drugs [31], enhance the transmission of neuromuscular junction [32] spreading from the brain to thorax will be done and we will test and the conduction between sensory afferents and motor neurons other drugs such as GABA, histamine and glutamate in [33], modulate the responses of proprioceptive sensory neurones modulating the electrically elicited flight behavior. [34], promote releasing of the peptidergic adipokinetic hormones As a preliminary study, this work has established an exper- [30], and regulate energy metabolism at the onset of flight [35]. imental system in honeybees and tested the behavioral responses to But in addition to octopamine, researches also have reported other stimulation of different brain subregions. In conclusion, this work neurotransmitters, neurohormones or neuromodulators on flight has helped us to understand which functional divisions of the modulation. Claassen and Kammer (1986) have demonstrated insect brain participate in flight control, and will support further that dopamine, octopamine and 5-HT are involved in initiating, studies to uncover the involved neurons from specific brain areas maintaining and terminating flight behavior, respectively [36]. via a neurophysiological approach, and also to test the hypothesis Brembs et al. (2007) show that octopamine and tyramine are regarding of the possible memory processing. involved in regulating flight initiation and maintenance through different sites, and therefore exert distinct effects on the flight Supporting Information central pattern generating network [37]. The methods that 28 Figure S1 Screenshots of two adjacent video frames pressure injecting biogenic amines or antagonists (10 – 210 displayed in 1/25 second by software Corel VideoStudio 10 mol) into thoracic ganglia or superfusing drug solutions Pro X5. Honeybee flight initiates in the second screenshot which directly onto the surface of flight muscles or thoracic ganglia were is distinguished by the high wing-beat frequency. used as a normal tool by researchers in such studies [36,38,39]. (TIF) Whereas we used an easier way by dripping diluted drug solutions (5610211 mol) directly onto the brain surface to obtain drug Figure S2 A curve chart displays the medio-lateral permeation. Though we find this method is not capable of positioning data of five brain subregions obtained from precluding the drug flow into thorax through haemolymph one honeybee brain frontal sections. Numbers on horizontal circulation. But from the results, we can conclude that only coordinate represent the measured brain slices from front to back. octopamine (out of octopamine, dopamine, serotonin and And the vertical coordinate shows the positioning data. Curves in acetylcholine) involves in the regulation of electrically elicited the color of green, purple, dark blue, orange and light blue

PLOS ONE | www.plosone.org 9 November 2014 | Volume 9 | Issue 11 | e113012 Flight Control in Honeybees represent the brain subregion of ellipsoid body, a-lobe, antennal Acknowledgments lobe, lobula and medulla respectively. Triangles on each curve represent the subregion centers and verticle lines indicate the We greatly appreciate the valuable comments and suggestions from the anonymous reviewers to improve this manuscript. Also, we thank Prof. subregion’s extended diameter. Jochen Zeil from ARC Centre of Excellence in Vision Science, the (TIF) Australian National University and Youfa Zhu from Core Facilities, Zhejiang University School of medicine for their help on honeybee brain Video S1 Induced flight by one stimulus on a-lobe in histology. We also would like to thank Shen Wang and Songchao Guo restrained honeybee. from the Qiushi Academy for Advanced Studies, Zhejiang University for (AVI) contributing to the design of honeybee clamping device. Video S2 Induced flight by three stimuli on medulla in restrained honeybee. Author Contributions (AVI) Conceived and designed the experiments: NGZ HXZ. Performed the experiments: HXZ NGZ LX FG. Analyzed the data: HXZ HQZ. Contributed reagents/materials/analysis tools: FLH XXZ WAR. Wrote the paper: HXZ NGZ WAR.

References 1. Ansari SA, Zbikowski R, Knowles K (2006) Aerodynamic modelling of insect- neuroprosthetic probe. in Proceedings of IEEE International Conference on like flapping flight for micro air vehicles. Progress in Aerospace Science 42(2): Micro Electro Mechanical System (MEMS 2010), Hong Kong, pp. 39–42. 129–172. 21. Bao L, Zheng N, Zhao H, Hao Y, Zheng H, et al. (2011) Flight control of 2. Pines DJ, Bohorquez F (2006) Challenges facing future micro-air-vehicle tethered honeybees using neural electrical stimulation. The 5th International development. Journal of Aircraft 43(2): 290–305. IEEE/EMBS Conference on Neural Engineering (NER), Cancun, Mexico, pp. 3. Zuo DC, Chen WY, Peng SL, Zhang WP (2006) Modeling and simulation study 558–561. of an insect-like flapping-wing micro aerial vehicle. Advanced Robotics 20(7): 22. Augustin H, Allen MJ, Partridge L (2011) Electrophysiological recordings from 807–824. the giant fiber pathway of D. melanogaster. Journal of Visualized Experiments 4. Tsang WM, Stone AL, Otten D, Aldworth ZN, Daniel TL, et al. (2012) Insect– 47: 2412, p. 5. machine interface: a carbon nanotube-enhanced flexible neural probe. Journal 23. Ejaz N, Peterson KD, Krapp HG (2011) An experimental platform to study the of Neuroscience Methods 204(2): 355–365. closed-loop performance of brain–machine interfaces. Journal of Visualized 5. Sato H, Maharbiz MM (2010) Recent developments in the remote radio control Experiments 49: 1677, p. 11. of insect flight. Frontiers in Neuroscience 4, p. 12. doi: 10.3389/fnins. 24. Tehovnik EJ (1996) Electrical stimulation of neural tissue to evoke behavioral 2010.00199. responses. Journal of Neuroscience Methods 65(1): 1–17. 6. Huber F (1960) Untersuchungen u¨ber die Funktion des Zentralnervensystems 25. Goodman LJ, Fletcher WA, Guy RG, Mobbs PG, Pomfrett CJD (1987) Motion und insbesondere des Gehirns bei der Fortbewegung und der Lauterzeugung der sensitive descending interneurons, ocellar LD neurons and neck motor neurons Grillen. Zeitschrift fu¨r vergleichende Physiologie 44: 60–132. in the bee: a neural substrate for visual course control in Apis mellifera.In 7. Otto D (1967) Untersuchungen zur nervo¨sen Kontrolle des Grillengesanges. Neurobiology and Behaviour of Honeybees (ed. R. Menzel and A. Mercer), pp. Zoologischer Anzeiger Supplement 31: 585–592. 158–171. Berlin, Heidelberg, New York: Springer-Verlag. 8. Otto D (1971) Untersuchungen zur zentralnervo¨sen Kontrolle der Lauterzeu- 26. Ibbotson MR, Goodman LJ (1990) Response characteristics of four wide-field gung von Grillen. Zeitschrift fu¨r vergleichende Physiologie 74: 227–271. motion-sensitive descending interneurones in Apis mellifera.Journalof 9. Bentley DR (1977) Control of cricket song patterns by descending interneurons. Experimental Biology 148: 255–279. Journal of Comparative Physiology A 116: 19–38. 27. Ibbotson MR (1991) Wide-field motion-sensitive neurons tuned to horizontal movement in the honeybee, Apis mellifere. Journal of Comparative Physiology A 10. Rowell CHF (1963) A method for chronically implanting stimulating electrodes 168: 91–102. into the brains of locusts, and some results of stimulation. Journal of 28. Oleskevich S, Clements JD, Srinivasan MV (1997) Long-term synaptic plasticity Experimental Biology 40: 271–284. in the honeybee. Journal of Neurophysiology 78: 528–532. 11. Blondeau J (1981) Electrically evoked course control in the fly Calliphora 29. Menzel R, Manz G (2005) Neural plasticity of mushroom body-extrinsic neurons . Journal of Experimental Biology 92: 143–153. erythrocephala in the honeybee brain. Journal of Experimental Biology 208: 4317–4332. 12. Holzer R, Shimoyama I (1997) Locomotion control of a bio-robotic system via 30. Orchard I, Ramirez JM, Lange AB (1993) A multifunctional role for octopamine electric stimulation. Proceedings of the International Conference on Intelligent in locust flight. Annual Review of Entomology 38: 227–249. Robots and Systems, Grenoble, France (1997), pp. 1514–1519. 31. Ramirez JM, Pearson KG (1991) Octopaminergic modulation of plateau 13. Sato H, Berry CW, Casey BE, Lavella G, Yao Y, et al. (2008a) A cyborg beetle: potentials in the flight system of the locust. Brain Research 549: 332–337. insect flight control through an implantable, tetherless microsystem. in 32. Evans PD, O’Shea M (1979) An octopaminergic neuronemodulates neuromus- Proceedings on Micro Electro Mechanical Systems (MEMS 2008), Tucson, cular transmission in the locust. Nature 270: 257–259. AZ, pp. 164–167. 33. Leitch B, Judge S, Pitman RM (2003) Octopaminergic modulation of synaptic 14. Sato H, Berry CW, Maharbiz MM (2008b) Flight control of 10 gram insects by transmission between an identified sensory afferent and flight motoneuron in the implanted neural stimulators. in Solid State Sensor Actuator Workshop, Hilton locust. Journal of Comparative Neurology 462: 55–70. Head Island, pp. 90–91. 34. Matheson T (1997) Octopamine modulates the responses and presynaptic 15. Sato H, Peeri Y, Baghoomian E, Berry CW, Maharbiz MM (2009a) Radio- inhibition of proprioceptive sensory neurones in the locust Schistocerca gregaria. controlled cyborg beetles: A radio-frequency system for insect neural flight Journal of Experimental Biology 200: 1317–1325. control. in Proceedings of the IEEE International Conference on Micro Electro 35. Mentel T, Duch C, Stypa H, Wegener G, Mueller U, et al. (2003) Central Mechanical Systems (MEMS 2009), Sorrento, Italy, pp. 216–219. modulatory neurons control fuel selection in flight muscle of migratory locust. 16. Sato H, Berry CW, Peeri Y, Baghoomian E, Casay BE, et al. (2009b) Remote Journal of Neuroscience 23: 1109–1113. radio control of insect flight. Frontiers in Integrative Neuroscience 3, p. 11. doi: 36. Claassen DE, Kammer AE (1986) Effects of octopamine, dopamine and 10.3389/neuro.07.024.2009. serotonin on production of flight motor output by thoracic ganglia of Manduca 17. Bozkurt A, Gilmour RF, Lal A (2009) Balloon-assisted flight of radio-controlled sexta. Journal of Neurobiology 17(1): 1–14. insect biobots. IEEE Transactions on Biomedical Engineering 56(9): 2304–2307. 37. Brembs B, Christiansen F, Pflu¨ger HJ, Duch C (2007) Flight initiation and 18. Tsang WM, Aldworth Z, Stone A, Permar A, Levine R, et al. (2008) Insect flight maintenance deficits in flies with genetically altered biogenic amine levels. control by neural stimulation of pupae-implanted flexible multisite electrodes. Journal of Neuroscience 27(41): 11122–11131. Proceeding of mTAS 2008, San Diego, USA, pp. 1922–1924. 38. Whim MD, Evans PD (1988) Octopaminergic modulation of flight muscle in the 19. Tsang WM, Stone AL, Aldworth ZN, Hildebrand JG, Daniel TL, et al. (2010a) locust. Journal of Experimental Biology 134: 247–266. Flexible split-ring electrode for insect flight biasing using multisite neural 39. Goldstein RS, Camhi JM (1991) Different effects of the biogenic amines stimulation. IEEE Transactions on Biomedical Engeening 57(7): 1757–1764. dopamine, serotonin and octopamine on the thoracic and abdominal portions of 20. Tsang WM, Stone AL, Aldworth ZN, Otten D, Akinwande AI, et al. (2010b) the escape circuit in the cockroach. Journal of Comparative Physiology A 168: Remote control of a cyborg moth using carbon nanotube-enhanced flexible 103–112.

PLOS ONE | www.plosone.org 10 November 2014 | Volume 9 | Issue 11 | e113012